Robust feature selection method for music classification

Human latent of distinguish varied music natures and cluster those into classes of categories are so incredible which expert in music will achieve such categorisation using their logical judgement and hearing senses. Till now, the technical society have concerned in delve into computerize the human way of distinguish the music in view of each necessary factor of the music tune, songs from voice of the artists are all based on the instrument types. The outcomes of those works up to now have been significant however still it can be enhanced. The main intend of this paper is to categorize the western and carnatic songs process automation by extracting the selected features which gives more accuracy of categorization. Inspite of the several features of audio signals (Method-1) the proposed method identifies the important robust features (Method-2) (RMS, Low Energy, Rolloff, Brightness, Pitch, Inharmonicity, Flatness and Entropy) with respect to carnatic/western music categorization. The identified features for carnatic/western music classification were tested using several classifiers and their accuracy of classification was discussed.